OPAL performance cookbook

Working directly in OPAL allows a wider range of options when modeling data. The following recommendations may give you better performance from your OPAL pipelines.

Optimizing Observe Performance

• Use approximate values when feasible
• Avoid large JSON blobs
• Cast data columns extracted from JSON
• Create intermediate Datasets
• Filter earlier in OPAL scripts
• Use filter instead of ever
• Flatten less first
• Limit Worksheet time windows
• Limit resource time windows
• Limit valid event time windows
• Look for hidden columns
• Use make_events before window functions
• Mark immutable resource columns
• Make resources from multiple Datasets
• Prefer join to lookup
• Prefer lead and lag to first and fast
• Prefer timechart to timestats
• Limit query time windows
• Define stricter time filters in queries
• Reduce columns earlier in OPAL scripts
• Extract from JSON instead of using flatten
• Type data columns
• Use interval for ephemeral things